Automated generation of discrete event controllers for dynamic reconfiguration of autonomous sensor networks

Abstract

Autonomous Sensor Networks have the potential for broad applicability to national security, intelligent transportation, industrial production and environmental and hazardous process control. Distributed sensors may be used for detecting bio-terrorist attacks, for contraband interdiction, border patrol, monitoring building safety and security, battlefield surveillance, or may be embedded in complex dynamic systems for enabling fault tolerant operations. In this paper we present algorithms and automation tools for constructing discrete event controllers for complex networked systems that restrict the dynamic behavior of the system according to given specifications. In our previous work we have modeled dynamic system as a discrete event automation whose open loop behavior is represented as a language <i>L</i> of strings generated with the alphabet 'Elipson' of all possible atomic events that cause state transitions in the network. The controlled behavior is represented by a sublanguage <i>K</i>, contained in <i>L</i>, that restricts the behavior of the system according to the specifications of the controller. We have developed the algebraic structure of controllable sublanguages as perfect right partial ideals that satisfy a precontrollability condition. In this paper we develop an iterative algorithm to take an <i>ad hoc </i>specification described using a natural language, and to formulate a complete specification that results in a controllable sublanguage. A supervisory controller modeled as an automaton that runs synchronously with the open loop system in the sense of Ramadge and Wonham is automatically generated to restrict the behavior of the open loop system to the controllable sublanguage. A battlefield surveillance scenario illustrates the iterative evolution of <i>ad hoc </i>specifications for controlling an autonomous sensor network and the generation of a controller that reconfigures the sensor network to dynamically adapt to environmental perturbations.